API Documentation

Quickstart

Calling models through Const Router requires only two changes to standard client code: point the base URL to https://router.const.web.id/v1 and supply your Const API key.

from openai import OpenAI

client = OpenAI(
    base_url="https://router.const.web.id/v1",
    api_key="const_live_YOUR_KEY",
)

response = client.chat.completions.create(
    model="deepseek-v4.1-flash",
    messages=[{"role": "user", "content": "Explain Virtual Combos in two sentences."}],
)
print(response.choices[0].message.content)

Authentication

Every inference request requires an active API key passed in the Authorization header with the Bearer prefix, or via x-api-key.

Keys are revealed once upon creation. Invalid, disabled, or daily-capped keys are rejected with HTTP 401/403 before executing upstream model calls.

Models & Free Allowance

Const Router operates two billing tiers: Free Tier and Paid Wholesale.

SDKs & Agent Setup

Use Const Router with official developer CLIs and autonomous agent frameworks:

export ANTHROPIC_BASE_URL="https://router.const.web.id"
export ANTHROPIC_API_KEY="const_live_YOUR_KEY"
claude

Chat Completions

Standard endpoint: POST /v1/chat/completions. Supports streaming, function/tool calling, structured JSON output, vision inputs, and reasoning controls.

curl https://router.const.web.id/v1/chat/completions \
  -H "Authorization: Bearer const_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash",
    "messages": [
      {"role": "system", "content": "You are a concise engineering assistant."},
      {"role": "user", "content": "Compare quicksort vs mergesort cache locality."}
    ],
    "temperature": 0.2,
    "max_tokens": 1024
  }'

For reasoning and thinking models (such as DeepSeek R1, GLM thinking, and OpenAI reasoning architectures), pass the reasoning_effort parameter to modulate internal thought token budgets before producing output.

from openai import OpenAI

client = OpenAI(
    base_url="https://router.const.web.id/v1",
    api_key="const_live_YOUR_KEY",
)

# Pass reasoning_effort parameter directly
response = client.chat.completions.create(
    model="deepseek-v4.1-flash",
    messages=[{"role": "user", "content": "Design an idempotent distributed lock with redis."}],
    reasoning_effort="high",
)

print(response.choices[0].message.content)

Streaming (Server-Sent Events)

Enable SSE by setting stream: true. Stream frames are passed through directly from upstream without proxy buffering delays. The final terminal chunk delivers exact token telemetry:

data: {"choices": [{"delta": {"content": "..."}}]}
...
data: {"usage": {"prompt_tokens": 42, "completion_tokens": 128}}
data: [DONE]

Tool Calling & JSON Mode

Const Router supports OpenAI standard function schemas via tools and tool_choice. For structured JSON outputs without tools, pass:

"response_format": { "type": "json_object" }

Virtual Combos & Automatic Failover

Virtual Combos allow grouping multiple upstream models under a single slug (model: "combo/my-coding-stack"). When your primary model returns HTTP 429 (rate limited), 503 (out of capacity), or high latency, Const Router automatically fails over to the next candidate model in the chain before replying to the client.

Available Endpoints

Errors & Retries

Standard HTTP error codes are preserved:

Per-Token Metering & Ledger

Every settled request emits an X-Const-Cost header with the exact micro-cost in USD deducted from your balance. Failed upstream requests are never billed.